基于视频图像的汽车驾驶员疲劳检测

发布时间：2018-08-24 20:57

【摘要】：汽车驾驶员疲劳检测分为接触式检测与非接触式检测,接触式测量就是直接接触驾驶员的身体,采集驾驶员的一些生理指标,通过这些指标来判断驾驶员的疲劳状态,非接触测量是指用汽车本身的某些参数或驾驶员的视频图像来判断驾驶员的疲劳状态。本文采用了基于视频检测的非接触式检测的方法。比起传统的接触检测方法,这种方法不需要接触驾驶员的身体,保障了行车过程中的安全和舒适。本文研究的内容包括:图像预处理、人脸检测、人眼检测,人眼轮廓提取,人眼轮廓的几何描述以及参数提取,人眼疲劳状态的判断。图像预处理过程中包括图像的灰度化、图像的去噪和图像的直方图均衡化,目的是减少图像的信息量使图像和排除图像中的噪声,使得图像易于处理。人脸检测和人眼检测都采用了Ada Boost分类器。用这种分类器可以构成一种级联分类,通过一个个逐级分类器最后得到分类结果,分类器选用了Haar-like特征。在检测人眼时,使用了两级的Ada Boost分类器,第一级分类器先检测出人脸,第二级分类器根据人脸和人眼的位置关系检测出了人眼,提高了人眼的检测率。人眼轮廓提取采用了数学形态学方法,人眼轮廓的几何描述采用了椭圆傅里叶变换法,这种变换将人眼的轮廓描述为傅里叶级数的系数,实验验证用很少级的傅里叶系数就可以准确地描述出人眼。PERCLOS参数反映的的是人眼在单位时间内的眨眼次数,它可以用来判断人眼的疲劳状态。最后,本文对驾驶员疲劳状态进行了仿真,结果达到了预期效果。
[Abstract]:Vehicle driver fatigue detection is divided into contact detection and non-contact detection. Contact measurement is to directly contact the driver's body, collect some physiological indicators of the driver, and judge the driver's fatigue state through these indicators. Non-contact measurement is used to judge the fatigue state of the driver by using some parameters of the vehicle or the video image of the driver. The method of contactless detection based on video detection is used in this paper. Compared with the traditional contact detection method, this method does not need to touch the driver's body, and ensures the safety and comfort of the driving process. The contents of this paper include: image preprocessing, face detection, human eye contour extraction, geometric description and parameter extraction of human eye contour, and judgment of human eye fatigue state. Image preprocessing includes grayscale image denoising and histogram equalization in order to reduce the amount of information in the image and eliminate the noise in the image making the image easy to process. Ada Boost classifier is used in face detection and human eye detection. This classifier can be used to form a cascade classification. Finally, the classification results are obtained by successive classifiers, and the classifier selects Haar-like features. In the detection of human eyes, a two-level Ada Boost classifier is used. The first level classifier detects the human face first, and the second level classifier detects the human eye according to the position relationship between the human face and the human eye, which improves the detection rate of the human eye. The human eye contour is extracted by mathematical morphology, and the geometric description of human eye contour is described by elliptical Fourier transform, which describes the contour of human eye as the coefficient of Fourier series. The experimental results show that the parameters of .PerCLOS can accurately describe the blink times of the human eye in a unit time, and it can be used to judge the fatigue state of the human eye. Finally, the fatigue state of the driver is simulated and the expected results are obtained.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41

【相似文献】